Deep Neural Network Solution for Detecting Intrusion in Network

Abstract

In our experiment, we found that deep learning surpassed machine learning when utilizing the DSSTE algorithm to sample imbalanced training set samples. These methods excel in terms of throughput due to their complex structure and ability to autonomously acquire relevant features from a dataset. The current study focuses on employing deep learning techniques such as RNN and Deep-NN, as well as algorithm design, to aid network IDS designers. Since public datasets already preprocess the data features, deep learning is unable to leverage its automatic feature extraction capability, limiting its ability to learn from preprocessed features. To harness the advantages of deep learning in feature extraction, mitigate the impact of imbalanced data, and enhance classification accuracy, our approach involves directly applying the deep learning model for feature extraction and model training on the existing network traffic data. By doing so, we aim to capitalize on deep learning's benefits, improving feature extraction, reducing the influence of imbalanced data, and enhancing classification accuracy.